Figure 4.

This figure illustrates how decision criteria and the old and new item distributions of evidence interact in the model to produce predictions. The bold lines and the bold text represent the changes in the values or predictions from the values in Panel A. In Panel A, the standard deviations in the across-trial memory distributions (i.e., the distributions of the means of the single-item evidence distributions) are equal and so are the decision criteria. The z-ROC function is linear with slope = 1. In Panels B and C the decision criteria are not equal. This produces a decrease in the slope of the z-ROC function. Panel D shows the effect of increasing the standard deviation in match across trials for old items, which produces a decrease in z-ROC slope. Note that the slope is not the ratio of the across trial distribution standard deviations. Panels E and F show how nonlinear z-ROC functions can be produced by altering decision criteria settings. For the confidence categories, “−−−” is high-confidence new, “+++” is high-confidence old, and the other symbols represent lower confidence responses. μ_o and μ_n are the means of the old and new across-item memory evidence distributions; σ_o and σ_n are their standard deviations. Z_Hit = z-score of the hit rate; Z_FA = z-score of the false alarm rate.